A color image restoration with adjacent parallel lines inhibition

We focus our attention on the problem of restoring a color image corrupted by blur and noise. This problem is ill-posed in the Hadamard sense. By regularization techniques, it is possible to define the solution of the problem as the minimum of an energy function. To improve the quality of the reconstruction, some Boolean variables, associated with the discontinuities of the intensity color field, are introduced. Such variables, called line variables, have to be estimated and can be used in the energy function in an implicit way. Moreover, we impose that adjacent discontinuities have not to be parallel. To minimize the energy function, a CATILED (convex approximation technique for interacting line elements deblurring) algorithm is used. The experimental results confirm the good qualities of this technique.